1 package jalview.ext.ensembl;
3 import jalview.analysis.AlignmentUtils;
4 import jalview.analysis.Dna;
5 import jalview.datamodel.Alignment;
6 import jalview.datamodel.AlignmentI;
7 import jalview.datamodel.DBRefEntry;
8 import jalview.datamodel.DBRefSource;
9 import jalview.datamodel.Mapping;
10 import jalview.datamodel.SequenceFeature;
11 import jalview.datamodel.SequenceI;
12 import jalview.exceptions.JalviewException;
13 import jalview.io.FastaFile;
14 import jalview.io.FileParse;
15 import jalview.io.gff.SequenceOntologyFactory;
16 import jalview.io.gff.SequenceOntologyI;
17 import jalview.util.DBRefUtils;
18 import jalview.util.MapList;
20 import java.io.IOException;
21 import java.net.MalformedURLException;
23 import java.util.ArrayList;
24 import java.util.Arrays;
25 import java.util.Collections;
26 import java.util.Comparator;
27 import java.util.List;
30 * Base class for Ensembl sequence fetchers
32 * @see http://rest.ensembl.org/documentation/info/sequence_id
35 public abstract class EnsemblSeqProxy extends EnsemblRestClient
37 private static final String ALLELES = "alleles";
39 protected static final String PARENT = "Parent";
41 protected static final String ID = "ID";
43 protected static final String NAME = "Name";
45 protected static final String DESCRIPTION = "description";
48 * enum for 'type' parameter to the /sequence REST service
50 public enum EnsemblSeqType
53 * type=genomic to fetch full dna including introns
58 * type=cdna to fetch coding dna including UTRs
63 * type=cds to fetch coding dna excluding UTRs
68 * type=protein to fetch peptide product sequence
73 * the value of the 'type' parameter to fetch this version of
78 EnsemblSeqType(String t)
83 public String getType()
91 * A comparator to sort ranges into ascending start position order
93 private class RangeSorter implements Comparator<int[]>
97 RangeSorter(boolean forward)
103 public int compare(int[] o1, int[] o2)
105 return (forwards ? 1 : -1) * Integer.compare(o1[0], o2[0]);
111 * Default constructor (to use rest.ensembl.org)
113 public EnsemblSeqProxy()
119 * Constructor given the target domain to fetch data from
121 public EnsemblSeqProxy(String d)
127 * Makes the sequence queries to Ensembl's REST service and returns an
128 * alignment consisting of the returned sequences.
131 public AlignmentI getSequenceRecords(String query) throws Exception
133 // TODO use a String... query vararg instead?
135 // danger: accession separator used as a regex here, a string elsewhere
136 // in this case it is ok (it is just a space), but (e.g.) '\' would not be
137 List<String> allIds = Arrays.asList(query
138 .split(getAccessionSeparator()));
139 AlignmentI alignment = null;
143 * execute queries, if necessary in batches of the
144 * maximum allowed number of ids
146 int maxQueryCount = getMaximumQueryCount();
147 for (int v = 0, vSize = allIds.size(); v < vSize; v += maxQueryCount)
149 int p = Math.min(vSize, v + maxQueryCount);
150 List<String> ids = allIds.subList(v, p);
153 alignment = fetchSequences(ids, alignment);
154 } catch (Throwable r)
157 String msg = "Aborting ID retrieval after " + v
158 + " chunks. Unexpected problem (" + r.getLocalizedMessage()
160 System.err.println(msg);
165 if (alignment == null)
171 * fetch and transfer genomic sequence features,
172 * fetch protein product and add as cross-reference
174 for (String accId : allIds)
176 addFeaturesAndProduct(accId, alignment);
179 for (SequenceI seq : alignment.getSequences())
181 getCrossReferences(seq);
188 * Fetches Ensembl features using the /overlap REST endpoint, and adds them to
189 * the sequence in the alignment. Also fetches the protein product, maps it
190 * from the CDS features of the sequence, and saves it as a cross-reference of
196 protected void addFeaturesAndProduct(String accId, AlignmentI alignment)
198 if (alignment == null)
206 * get 'dummy' genomic sequence with exon, cds and variation features
208 SequenceI genomicSequence = null;
209 EnsemblFeatures gffFetcher = new EnsemblFeatures(getDomain());
210 EnsemblFeatureType[] features = getFeaturesToFetch();
211 AlignmentI geneFeatures = gffFetcher.getSequenceRecords(accId,
213 if (geneFeatures.getHeight() > 0)
215 genomicSequence = geneFeatures.getSequenceAt(0);
217 if (genomicSequence != null)
220 * transfer features to the query sequence
222 SequenceI querySeq = alignment.findName(accId);
223 if (transferFeatures(accId, genomicSequence, querySeq))
227 * fetch and map protein product, and add it as a cross-reference
228 * of the retrieved sequence
230 addProteinProduct(querySeq);
233 } catch (IOException e)
235 System.err.println("Error transferring Ensembl features: "
241 * Returns those sequence feature types to fetch from Ensembl. We may want
242 * features either because they are of interest to the user, or as means to
243 * identify the locations of the sequence on the genomic sequence (CDS
244 * features identify CDS, exon features identify cDNA etc).
248 protected abstract EnsemblFeatureType[] getFeaturesToFetch();
251 * Fetches and maps the protein product, and adds it as a cross-reference of
252 * the retrieved sequence
254 protected void addProteinProduct(SequenceI querySeq)
256 String accId = querySeq.getName();
259 AlignmentI protein = new EnsemblProtein(getDomain())
260 .getSequenceRecords(accId);
261 if (protein == null || protein.getHeight() == 0)
263 System.out.println("No protein product found for " + accId);
266 SequenceI proteinSeq = protein.getSequenceAt(0);
269 * need dataset sequences (to be the subject of mappings)
271 proteinSeq.createDatasetSequence();
272 querySeq.createDatasetSequence();
274 MapList mapList = AlignmentUtils.mapCdsToProtein(querySeq, proteinSeq);
277 // clunky: ensure Uniprot xref if we have one is on mapped sequence
278 SequenceI ds = proteinSeq.getDatasetSequence();
279 ds.setSourceDBRef(proteinSeq.getSourceDBRef());
281 Mapping map = new Mapping(ds, mapList);
282 DBRefEntry dbr = new DBRefEntry(getDbSource(), getDbVersion(),
283 proteinSeq.getName(), map);
284 querySeq.getDatasetSequence().addDBRef(dbr);
287 * copy exon features to protein, compute peptide variants from dna
288 * variants and add as features on the protein sequence ta-da
290 AlignmentUtils.computeProteinFeatures(querySeq, proteinSeq, mapList);
292 } catch (Exception e)
295 .println(String.format("Error retrieving protein for %s: %s",
296 accId, e.getMessage()));
301 * Get database xrefs from Ensembl, and attach them to the sequence
305 protected void getCrossReferences(SequenceI seq)
307 while (seq.getDatasetSequence() != null)
309 seq = seq.getDatasetSequence();
312 EnsemblXref xrefFetcher = new EnsemblXref(getDomain());
313 List<DBRefEntry> xrefs = xrefFetcher.getCrossReferences(seq.getName());
314 for (DBRefEntry xref : xrefs)
318 * Save any Uniprot xref to be the reference for SIFTS mapping
320 if (DBRefSource.UNIPROT.equals(xref.getSource()))
322 seq.setSourceDBRef(xref);
327 * and add a reference to itself
329 DBRefEntry self = new DBRefEntry(getDbSource(), "0", seq.getName());
334 * Fetches sequences for the list of accession ids and adds them to the
335 * alignment. Returns the extended (or created) alignment.
340 * @throws JalviewException
341 * @throws IOException
343 protected AlignmentI fetchSequences(List<String> ids, AlignmentI alignment)
344 throws JalviewException, IOException
346 if (!isEnsemblAvailable())
349 throw new JalviewException("ENSEMBL Rest API not available.");
351 FileParse fp = getSequenceReader(ids);
352 FastaFile fr = new FastaFile(fp);
353 if (fr.hasWarningMessage())
355 System.out.println(String.format(
356 "Warning when retrieving %d ids %s\n%s", ids.size(),
357 ids.toString(), fr.getWarningMessage()));
359 else if (fr.getSeqs().size() != ids.size())
361 System.out.println(String.format(
362 "Only retrieved %d sequences for %d query strings", fr
363 .getSeqs().size(), ids.size()));
366 if (fr.getSeqs().size() == 1 && fr.getSeqs().get(0).getLength() == 0)
369 * POST request has returned an empty FASTA file e.g. for invalid id
371 throw new IOException("No data returned for " + ids);
374 if (fr.getSeqs().size() > 0)
376 AlignmentI seqal = new Alignment(
377 fr.getSeqsAsArray());
378 for (SequenceI sq:seqal.getSequences())
380 if (sq.getDescription() == null)
382 sq.setDescription(getDbName());
384 String name = sq.getName();
385 if (ids.contains(name)
386 || ids.contains(name.replace("ENSP", "ENST")))
388 DBRefUtils.parseToDbRef(sq, DBRefSource.ENSEMBL, "0", name);
391 if (alignment == null)
397 alignment.append(seqal);
404 * Returns the URL for the REST call
407 * @throws MalformedURLException
410 protected URL getUrl(List<String> ids) throws MalformedURLException
413 * a single id is included in the URL path
414 * multiple ids go in the POST body instead
416 StringBuffer urlstring = new StringBuffer(128);
417 urlstring.append(getDomain() + "/sequence/id");
420 urlstring.append("/").append(ids.get(0));
422 // @see https://github.com/Ensembl/ensembl-rest/wiki/Output-formats
423 urlstring.append("?type=").append(getSourceEnsemblType().getType());
424 urlstring.append(("&Accept=text/x-fasta"));
426 URL url = new URL(urlstring.toString());
431 * A sequence/id POST request currently allows up to 50 queries
433 * @see http://rest.ensembl.org/documentation/info/sequence_id_post
436 public int getMaximumQueryCount()
442 protected boolean useGetRequest()
448 protected String getRequestMimeType(boolean multipleIds)
450 return multipleIds ? "application/json" : "text/x-fasta";
454 protected String getResponseMimeType()
456 return "text/x-fasta";
461 * @return the configured sequence return type for this source
463 protected abstract EnsemblSeqType getSourceEnsemblType();
466 * Returns a list of [start, end] genomic ranges corresponding to the sequence
469 * The correspondence between the frames of reference is made by locating
470 * those features on the genomic sequence which identify the retrieved
471 * sequence. Specifically
473 * <li>genomic sequence is identified by "transcript" features with
474 * ID=transcript:transcriptId</li>
475 * <li>cdna sequence is identified by "exon" features with
476 * Parent=transcript:transcriptId</li>
477 * <li>cds sequence is identified by "CDS" features with
478 * Parent=transcript:transcriptId</li>
481 * The returned ranges are sorted to run forwards (for positive strand) or
482 * backwards (for negative strand). Aborts and returns null if both positive
483 * and negative strand are found (this should not normally happen).
485 * @param sourceSequence
488 * the start position of the sequence we are mapping to
491 protected MapList getGenomicRangesFromFeatures(SequenceI sourceSequence,
492 String accId, int start)
494 SequenceFeature[] sfs = sourceSequence.getSequenceFeatures();
501 * generously initial size for number of cds regions
502 * (worst case titin Q8WZ42 has c. 313 exons)
504 List<int[]> regions = new ArrayList<int[]>(100);
505 int mappedLength = 0;
506 int direction = 1; // forward
507 boolean directionSet = false;
509 for (SequenceFeature sf : sfs)
512 * accept the target feature type or a specialisation of it
513 * (e.g. coding_exon for exon)
515 if (identifiesSequence(sf, accId))
517 int strand = sf.getStrand();
518 strand = strand == 0 ? 1 : strand; // treat unknown as forward
520 if (directionSet && strand != direction)
522 // abort - mix of forward and backward
523 System.err.println("Error: forward and backward strand for "
531 * add to CDS ranges, semi-sorted forwards/backwards
535 regions.add(0, new int[] { sf.getEnd(), sf.getBegin() });
539 regions.add(new int[] { sf.getBegin(), sf.getEnd() });
541 mappedLength += Math.abs(sf.getEnd() - sf.getBegin() + 1);
546 * 'gene' sequence is contiguous so we can stop as soon as its
547 * identifying feature has been found
554 if (regions.isEmpty())
556 System.out.println("Failed to identify target sequence for " + accId
557 + " from genomic features");
562 * a final sort is needed since Ensembl returns CDS sorted within source
563 * (havana / ensembl_havana)
565 Collections.sort(regions, new RangeSorter(direction == 1));
567 List<int[]> to = Arrays.asList(new int[] { start,
568 start + mappedLength - 1 });
570 return new MapList(regions, to, 1, 1);
574 * Answers true if the sequence being retrieved may occupy discontiguous
575 * regions on the genomic sequence.
577 protected boolean isSpliceable()
583 * Returns true if the sequence feature marks positions of the genomic
584 * sequence feature which are within the sequence being retrieved. For
585 * example, an 'exon' feature whose parent is the target transcript marks the
586 * cdna positions of the transcript.
592 protected abstract boolean identifiesSequence(SequenceFeature sf,
596 * Transfers the sequence feature to the target sequence, locating its start
597 * and end range based on the mapping. Features which do not overlap the
598 * target sequence are ignored.
601 * @param targetSequence
603 * mapping from the sequence feature's coordinates to the target
605 * @param forwardStrand
607 protected void transferFeature(SequenceFeature sf,
608 SequenceI targetSequence, MapList mapping, boolean forwardStrand)
610 int start = sf.getBegin();
611 int end = sf.getEnd();
612 int[] mappedRange = mapping.locateInTo(start, end);
614 if (mappedRange != null)
616 SequenceFeature copy = new SequenceFeature(sf);
617 copy.setBegin(Math.min(mappedRange[0], mappedRange[1]));
618 copy.setEnd(Math.max(mappedRange[0], mappedRange[1]));
619 targetSequence.addSequenceFeature(copy);
622 * for sequence_variant on reverse strand, have to convert the allele
623 * values to their complements
626 && SequenceOntologyFactory.getInstance().isA(sf.getType(),
627 SequenceOntologyI.SEQUENCE_VARIANT))
629 reverseComplementAlleles(copy);
635 * Change the 'alleles' value of a feature by converting to complementary
636 * bases, and also update the feature description to match
640 static void reverseComplementAlleles(SequenceFeature sf)
642 final String alleles = (String) sf.getValue(ALLELES);
647 StringBuilder complement = new StringBuilder(alleles.length());
648 for (String allele : alleles.split(","))
650 reverseComplementAllele(complement, allele);
652 String comp = complement.toString();
653 sf.setValue(ALLELES, comp);
654 sf.setDescription(comp);
657 * replace value of "alleles=" in sf.ATTRIBUTES as well
658 * so 'output as GFF' shows reverse complement alleles
660 String atts = sf.getAttributes();
663 atts = atts.replace(ALLELES + "=" + alleles, ALLELES + "=" + comp);
664 sf.setAttributes(atts);
669 * Makes the 'reverse complement' of the given allele and appends it to the
670 * buffer, after a comma separator if not the first
675 static void reverseComplementAllele(StringBuilder complement,
678 if (complement.length() > 0)
680 complement.append(",");
682 if ("HGMD_MUTATION".equalsIgnoreCase(allele))
684 complement.append(allele);
688 char[] alleles = allele.toCharArray();
689 for (int i = alleles.length - 1; i >= 0; i--)
691 complement.append(Dna.getComplement(alleles[i]));
697 * Transfers features from sourceSequence to targetSequence
700 * @param sourceSequence
701 * @param targetSequence
702 * @return true if any features were transferred, else false
704 protected boolean transferFeatures(String accessionId,
705 SequenceI sourceSequence, SequenceI targetSequence)
707 if (sourceSequence == null || targetSequence == null)
712 // long start = System.currentTimeMillis();
713 SequenceFeature[] sfs = sourceSequence.getSequenceFeatures();
714 MapList mapping = getGenomicRangesFromFeatures(sourceSequence, accessionId,
715 targetSequence.getStart());
721 boolean result = transferFeatures(sfs, targetSequence, mapping,
723 // System.out.println("transferFeatures (" + (sfs.length) + " --> "
724 // + targetSequence.getSequenceFeatures().length + ") to "
725 // + targetSequence.getName()
726 // + " took " + (System.currentTimeMillis() - start) + "ms");
731 * Transfer features to the target sequence. The start/end positions are
732 * converted using the mapping. Features which do not overlap are ignored.
733 * Features whose parent is not the specified identifier are also ignored.
736 * @param targetSequence
741 protected boolean transferFeatures(SequenceFeature[] features,
742 SequenceI targetSequence, MapList mapping, String parentId)
744 final boolean forwardStrand = mapping.isFromForwardStrand();
747 * sort features by start position (which corresponds to end
748 * position descending if reverse strand) so as to add them in
749 * 'forwards' order to the target sequence
751 sortFeatures(features, forwardStrand);
753 boolean transferred = false;
754 for (SequenceFeature sf : features)
756 if (retainFeature(sf, parentId))
758 transferFeature(sf, targetSequence, mapping, forwardStrand);
766 * Sort features by start position ascending (if on forward strand), or end
767 * position descending (if on reverse strand)
770 * @param forwardStrand
772 protected static void sortFeatures(SequenceFeature[] features,
773 final boolean forwardStrand)
775 Arrays.sort(features, new Comparator<SequenceFeature>()
778 public int compare(SequenceFeature o1, SequenceFeature o2)
782 return Integer.compare(o1.getBegin(), o2.getBegin());
786 return Integer.compare(o2.getEnd(), o1.getEnd());
793 * Answers true if the feature type is one we want to keep for the sequence.
794 * Some features are only retrieved in order to identify the sequence range,
795 * and may then be discarded as redundant information (e.g. "CDS" feature for
798 @SuppressWarnings("unused")
799 protected boolean retainFeature(SequenceFeature sf, String accessionId)
801 return true; // override as required
805 * Answers true if the feature has a Parent which refers to the given
806 * accession id, or if the feature has no parent. Answers false if the
807 * feature's Parent is for a different accession id.
813 protected boolean featureMayBelong(SequenceFeature sf, String identifier)
815 String parent = (String) sf.getValue(PARENT);
816 // using contains to allow for prefix "gene:", "transcript:" etc
817 if (parent != null && !parent.contains(identifier))
819 // this genomic feature belongs to a different transcript
826 public String getDescription()
828 return "Ensembl " + getSourceEnsemblType().getType()
829 + " sequence with variant features";
833 * Returns a (possibly empty) list of features on the sequence which have the
834 * specified sequence ontology type (or a sub-type of it), and the given
835 * identifier as parent
842 protected List<SequenceFeature> findFeatures(SequenceI sequence,
843 String type, String parentId)
845 List<SequenceFeature> result = new ArrayList<SequenceFeature>();
847 SequenceFeature[] sfs = sequence.getSequenceFeatures();
849 SequenceOntologyI so = SequenceOntologyFactory.getInstance();
850 for (SequenceFeature sf :sfs) {
851 if (so.isA(sf.getType(), type))
853 String parent = (String) sf.getValue(PARENT);
854 if (parent.equals(parentId))
865 * Answers true if the feature type is either 'NMD_transcript_variant' or
866 * 'transcript' or one of its sub-types in the Sequence Ontology. This is
867 * needed because NMD_transcript_variant behaves like 'transcript' in Ensembl
868 * although strictly speaking it is not (it is a sub-type of
874 public static boolean isTranscript(String featureType)
876 return SequenceOntologyI.NMD_TRANSCRIPT_VARIANT.equals(featureType)
877 || SequenceOntologyFactory.getInstance().isA(featureType,
878 SequenceOntologyI.TRANSCRIPT);